The Yale Law Journal

April 2023

The Accountable Bureaucrat

Administrative Law

abstract. Common wisdom has it that bureaucrats are unaccountable to the people they regulate and must therefore be closely supervised by elected officials or (perhaps ironically) the federal courts. For many detractors of the administrative state, as well as many proponents, agency accountability hangs on the concentrated power of the President in particular. This Article presents a different vision. Drawing on in-depth interviews with officials from numerous agencies, we show that everyday administrative practices and relationships themselves support accountability of a kind that neither elections nor judicial review alone can achieve.

Our interviews reveal that agency officials work within structures that promote the very values accountability is supposed to serve: deliberation, inclusivity, and responsiveness. Three primary features of the administrative state support this vision of accountability. First, political appointees and career civil servants, often presented as adversaries, actually represent complementary decision-making modalities. Appointees do not impose direct presidential control but imbue agencies with a diffuse, differentiated sense of abstract political values and policy priorities tied to the electoral and civil-society coalitions that support the administration in power. Civil servants use expertise and experience to set the parameters within which decisions can be made. Combining these different but interdependent approaches to policymaking promotes deliberation informed by public opinion and the public interest. Second, agencies work through what we refer to as a decision-making web, which facilitates continual justification and negotiation among officials with different roles inside the state. This claim stands in stark contrast to the strict hierarchy often attributed to government bureaucracy. We show how the principal-agent model gives way, more often than not, to the dispersion of decision-making power, which promotes the pluralistic inclusivity of views in a way hierarchical decision-making does not. Finally, numerous practices connect agencies directly and pervasively to the people and situations they regulate. Those required by law, like notice-and-comment rulemaking, are supplemented by varied other means by which agencies respond and adapt to the views of affected publics and the realities of the regulated world.

Our research provides crucial empirical evidence of how the everyday work of government gets done and gives the often-invoked notion of accountability some real content. It leads us to reject formalistic claims about what constitutes accountability in the abstract and to focus instead on the relationships, structures, and practices that actually promote accountability—features of the administrative state that help head off arbitrariness, incorporate multiple perspectives, and encourage negotiated, provisional outcomes. These resources for promoting republican democracy within the bureaucracy, however, are neither inherent nor eternal: they must be actively nourished. This Article thus should change how we think about government accountability and inform how we structure our institutions to achieve it.

authors. Anya Bernstein is Professor of Law, University of Connecticut School of Law. Substantial parts of this Article were written while she was a Visiting Scholar at the Max Planck Institute for Social Anthropology’s Department of Law and Anthropology. She thanks the Department and its director, Marie-Claire Foblets, for providing a supportive environment for intellectual pursuits, and Larissa Vetters for many helpful conversations. She is grateful for the generous support of the Fulbright Foundation, which supported that visit with a U.S. Fulbright Scholar award. Cristina Rodríguez is Leighton Homer Surbeck Professor of Law, Yale Law School. She is grateful for generous financial support from the Oscar M. Ruebhausen Fund at Yale Law School, which greatly assisted in the research supporting this Article. Both authors are deeply grateful for the excellent insights and feedback they received in response to this Article. They would like to thank Daniel Carpenter, Brian Cook, Brian Feinstein, David Fontana, Chris Havasy, Christine Jolls, Amy Kapczynski, Joshua Macey, Neysun Mahboubi, Jerry Mashaw, Alex Mechanick, Gillian Metzger, Jennifer Nou, Anne Joseph O’Connell, Nick Parrillo, Shalev Roisman, Roberta Romano, Susan Rose-Ackerman, Kate Shaw, Jodi Short, Reva Siegel, Norman Silber, David Spence, Glen Staszewski, Jed Stiglitz, John Witt, Ilan Wurman, and David Zaring, as well as participants in faculty workshops at Yale Law School, the University of Toronto Faculty of Law, the University of Connecticut Law School, and the Max Planck Institute for Social Anthropology, and those in the seminar on Power in the Administrative State. Special thanks to Errol Meidinger for encouraging this project in its earliest stages. This work also benefitted enormously from several generations of talented students who played essential roles in the conceptualization and execution of this multiyear project. For their dedication to the data gathering and coding phases of our project, the authors thank Ashraf Ahmed, Natasha Khan, Alexandra Lichtenstein, Alexander Mechanick, William Powell, Kshithij Shrinath, Becca Steinberg, and Helen White. For their excellent work in helping them design the project, they thank Anna Mohan and Megan McGlynn. And for their assistance in polishing this Article, they thank Colin Burke, David Hopen, Jim Huang, Beatrice Pollard, Chelsea Thomeer, Bardia Vaseghi, and Charlotte Witherspoon. The authors’ greatest thanks go to the numerous agency officials who generously shared their experiences and insights, not to mention their time, to make this research possible.


The image of the unelected bureaucrat indicts the administrative state. Judges and critics regularly marshal this menacing figure to challenge the excessive power and questionable legitimacy of the bureaucracy as a whole. On one view, the unelected bureaucrat heightens the dangers posed by executive government. When the Supreme Court enjoined the Biden Administration’s regulation requiring many employees to be vaccinated or test regularly for COVID-19, Justice Gorsuch praised the decision for helping “to prevent ‘government by bureaucracy supplanting government by the people.’”1 The unelected bureaucrat also threatens the rightful power of the President. Chief Justice Roberts, on behalf of Court majorities, has expressed fear that removal protections for certain officials can lead agencies to “slip from the Executive’s control, and thus from that of the people.”2 Whether aggrandizing the executive branch or sapping its one elected official of power, the figure of the unelected bureaucrat is a reliably winning card.3 Even many who support robust administration sometimes accept the premise that unelected bureaucrats pose a threat to accountable government. The dissent in the vaccine-or-test case, for example, argued that the agency’s rule had “the virtue of political accountability, for [the Occupational Safety and Health Administration (OSHA)] is responsible to the President, and the President is responsible to—and can be held to account by—the American public.”4

These claims all rest on the commonsense notion that elections create government accountability.5 In this Article, we argue instead that accountability within the American administrative state arises from elections only indirectly, if at all. Our empirical research, involving interviews with administrators across a range of federal agencies,6 reveals numerous structures, relationships, and practices within the state itself that produce various and important forms of accountability. These scaffolds augment and complement the accountability created by elections, in some cases producing the very sorts of accountability that elections supposedly, but do not actually, provide. Based on these findings, we argue that placing excessive emphasis on elections or elected officials—whether to constrict, justify, or structure administration—gets in the way of understanding how to build and sustain an accountable democratic state.

Democracy depends on accountability: those who exercise power should be held responsible for their actions and decisions.7 At its most basic, accountability requires government actors to justify their positions so that others can evaluate, challenge, or override them.8 Such justification has distinct payoffs. It renders arbitrary or biased views more visible and contestable and pushes government actors to consider multiple perspectives in their decision-making. It allows interested publics to test the quality of government decisions and to change them over time, giving democratic subjects an ongoing role in their own governance.9 And it ensures that public institutions serve as sites for the contestations, negotiations, and provisional outcomes that characterize any successful democracy.10 This understanding of accountability fits with the notion of republican democracy, which rejects the arbitrary exercise of power or power that fails to take relevant interests into account.11 In short, accountability occurs when government is made deliberative, inclusive, and responsive.

Agency action is often regarded by judges and commentators as accountable only insofar as it can be directly controlled by the elected President.12 But our research locates the production of accountability elsewhere. We identify three mutually supporting aspects of agency practice as particularly important to ensuring accountability: (1) diffuse, rather than concentrated, forms of political control; (2) nonhierarchical organizational structures of negotiation and deliberation among numerous actors and groups; and (3) practices that keep agencies attuned to affected publics and events in the regulated world. In this Article, we present each of these features of administration and explain how they should change the way scholars and lawyers conceptualize and pursue accountability in government.

In Part I, we show that, as a matter of practice, direct presidential involvement in agency policymaking is neither frequent nor the most important source of political control. Instead, political influence is usually broader and more diffuse.13 Political appointees chosen by the President or the heads of agencies play a crucial role, but they do not merely carry out the President’s agenda. These appointees are usually relatively independent, acting on policy orientations congruent with, but not in any clear sense directed by, the President. Their bigger contribution lies in introducing forms of reasoning and decision-making that complement the work of career officials. The complex relationships between these epistemic communities help structure agency accountability. Meanwhile, presidential administration, though now entrenched, is often less significant than either its proponents or its detractors claim.14 We show how the broader political connection we identify helps ground accountability not through a direct connection to elections but through interdependent relationships between different modalities of decision-making.

In Part II, we bring to light the networked structures through which agency policymaking proceeds. Commentators sometimes assume that accountability in the administrative state rests on clear hierarchies and fairly simple principal-agent relationships.15 We found agency action to be characterized by something quite different: broad participation, multifarious input, and ongoing reason-giving characterized as much by negotiation as by supervision.16 Participants in this process continually justify their positions not just to a particular principal but to many players within the bureaucracy. And supervisors’ specific policy preferences often do not precede but rather emerge through policymaking practices themselves. Our findings thus challenge the simplified principal-agent model of agency practice and suggest that accountability can emerge from dispersed, rather than consolidated, authority.

In Part III, we pan out to consider agencies’ external relations. Agencies frequently and intentionally react to events in the world and the public opinion that arises from those events. Officials utilize both the formal channels created through the notice-and-comment process and myriad informal, semistructured channels they themselves have opened to engage with the public.17 In so doing, agencies render themselves subject to evaluation and influence by those whom their decisions affect. Bureaucrats’ relative insulation from elections leads some critics to assume that they are walled off from the world—computers in sealed rooms. But lacking a direct electoral connection does not keep administrators removed from regulated entities or the regulated world. Our findings suggest that, on the contrary, agencies have more diverse, frequent, and interactive relationships with the publics and situations they regulate than elections could provide.18 These ongoing interactions and agencies’ attendant concerns with the efficacy of their actions promote accountability as well.

We conclude by considering the implications of the relational, negotiated, and contextual forms of accountability we identify. A popular challenge to the administrative state paints it as undemocratic—in the most sinister formulation, a “deep” state that threatens our freedoms.19 We find, in contrast, that everyday agency processes facilitate accountability. They build in requirements for officials to continually present ideas to be reviewed, vetted, and tested by other actors who bring to bear different forms of judgment and expertise and who themselves hold meaningful stakes in the policymaking process. Such interactions routinely lead to the reconsideration of views with respect to legal questions, implementation issues, and policy desirability. These practices provide scaffolding that makes it possible for participants to enact the virtues of accountability: to be pragmatically responsive to social needs, to problem-solve in the public interest, and to justify the exercise of government power. The features of agency decision-making that we uncover are thus worth pursuing as a matter of institutional design, and their potential presence should inform both legal doctrines and political perceptions related to bureaucracy.

Elections, of course, remain crucial to legitimating government,20 and their effects do and should permeate agency decision-making.21 But real accountability requires
more.22 Narrowing the notion of accountability to the electoral connection instantiates a peculiarly anemic notion of democracy that leaves out many of the traits that make democratic governance normatively attractive.23 Doing so amounts to insisting that elections do work for which they are not suited and ignores most of the accountability-forcing work that needs to be done. The legitimation produced through elections finds its complement in the practices we detail here, which build accountability into government work itself. In some sense, then, the imprecation that “unelected bureaucrats” are unaccountable is a complaint about the wrong thing: if bureaucrats are unaccountable, it is not because they are unelected but because they do not utilize the accountability structures that administration provides.24

Unfortunately, the Supreme Court has begun to severely constrain the administrative state in the name of this anemic conception of democracy. Suspicion of the unelected bureaucrat is used to justify limiting both how Congress can design the regulatory state25 and how agencies can address urgent problems in their statutory purview.26 The Justices support these limits in the name of democracy, but their reliance on a formalistic conception of democratic legitimacy, unmoored from actual practices, threatens to undermine, rather than strengthen, agency accountability.

Our analysis also departs from a common assumption that government accountability depends largely on strict hierarchy or principal-agent relationships. The simplified principal-agent model, the foundation of much doctrine and legal scholarship on the administrative state, cannot account for the wealth of mutual accounting we found cutting across institutional hierarchies: the peers who needed to justify their positions to one another, the political appointees who had to persuade career staff to move an idea forward. These relationships, even when strictly speaking hierarchical, often do not function on a command-and-control basis. The model also fails to capture the dynamic development not just of policies but of policy preferences. It is often not the case that a superior commands a subordinate to carry out a predetermined policy position. Such a position is just as likely something that develops in the policymaking process itself: deliberation shapes preferences.27 Of course, agencies provide many opportunities for superiors to give clear instructions that subordinates must obey. But taking the simplified principal-agent relationship as the model for government action obscures more than it reveals.

In addition to challenging some of the conventions of administrative law, our research contributes to a growing field of empirical studies of bureaucracy, which to date has focused largely on the “street-level bureaucrats” who make individualized determinations in a given policy context.28 We study instead those we might call elevator-level administrators—the ones who produce the policies to begin with.29 We do so not to explain how any particular policy was produced or how any specific agency functions, but rather to illuminate policymaking trends and mechanisms broadly operative in the administrative state.30 Given the variety of agency forms and cultures within the American bureaucracy, we do not claim that our picture is comprehensive. But our interviewees, dispersed across different roles in various agencies, overlapped sufficiently in their descriptions of agency practices to provide important empirical insight into the relationship between administrative policymaking and a well-functioning democracy.31

We do not claim that administrative agencies are accountable in some transhistorical, inherent way. Accountability inevitably depends on empirical realities that differ across circumstances: institutions can be more or less, and also differently, accountable depending on their participants, their structures, their cultures, and so on.32 Instead of asking whether an institution is accountable, we ask what scaffolds it provides to support and channel accountability. Such scaffolds do not themselves create accountability. There is always the possibility that people will shirk, undermine, or attack them.33 But these scaffolds encourage and enable accountability. Those we found in our work rest upon an assumption of pluralism, provide means for mediating differing interests, and build responsiveness to outside interests and events into agency action. Again, these findings do not mean that the bureaucracy will necessarily be accountable; but they do mean that it has developed tools for accountability—tools that we should understand, defend, and even replicate.34

While we do not claim to present a complete or eternal picture of government accountability, our study raises normative and political questions that discussions of accountability too often ignore: questions about why accountability matters, what purposes it serves, and what it should look like in practice. That is, we not only provide empirical insight into accountability practices but advance the theorization of what accountability entails. We show that the bureaucracy can be accountable even though the officials who comprise it are unelected. But we also resist overclaiming the democratic bona fides and efficacy of presidential control. Our research should thus change how we think about accountability in the state and inform how we structure institutions to achieve it.

* * *

Before we get to our findings, a word on our methodology.35 Much of the information and insight of this project comes from thirty-nine in-depth, open-ended interviews with political appointees and career civil servants from eleven agencies, small and large, old and new, benefit-managing and conduct-regulating. Interviews are useful for eliciting new information and unknown views because they solicit participant perspectives, experiences, and ideas. Our primary objective in conducting the interviews was to better understand agency statutory interpretation and policymaking, and we therefore inquired into interpretive approaches and tools; relations to Congress, the President, other agencies, and states; judicial influence; and participants’ views on how agencies work with statutes. We asked no questions about accountability. Our conclusion that the practices interviewees described promoted government accountability emerged from our engagement with the interview transcripts, not from our subjects’ self-descriptions.

In devising our interview protocols, we did not aim to produce structurally identical, easily comparable responses; we sought instead complex, nuanced images of administration from the bureaucrat’s point of view.36 To find resonances in these disparate conversations, we identified major categories of analysis in the interview transcripts and coded each passage with its relevant categories.37 This engagement with the material also yielded an ambient sense of how our subjects’ descriptions converged, which further guided our approach to the data.38 Data collection, coding, and interpretation were informed by the qualitative methods of anthropology, which seek to explain how people create social orders and meanings.39 Through this process, we recognized that our respondents described everyday work practices that corresponded to the values that scholarship and political discourse often demand of accountable democratic governance.

Like any method, ours provides only a partial view of the object of analysis. Our limited number of participants constrains the range of experiences we encompassed, though we find this limitation in breadth amply balanced by the depth of discussion elicited by our wide-ranging, lengthy interviews. The extent to which our institutionally diverse respondents converged on key parts of their work leaves us confident that we have identified some central aspects of our administrative state.